Offload pipeline variables to Object Storage

Problem

Partitions ci_pipeline_variables_* take a lot of space.

Proposal

Evaluate what we need to do to move this data to Object Storage.

Given the simple access patterns of this data:

• CRUD via project settings
• READ when building job variables (cached at pipeline level)

It would be ideal to store this data outside the database.

Details

• understand how we would store encrypted data at rest and decrypt on read
• understand if best to store all variables in a single blob on OS (ideal for cleanup and less db references)
• define necessary abstractions that could be reused for other data types.